AI framework for autonomous systems

The key feature of autonomous systems is that they use sensors to map their environment and can interact with it independently using actuators. For example, this paves the way for self-driving cars, robots that perform tasks autonomously, and systems that regulate themselves adaptively. Autonomous systems are made up of sensors for mapping the environment and components for the aggregation, analysis, and interpretation of data, as well as situation assessment, action planning, and actuators. A method known as deep reinforcement learning (DRL) is used to implement decision-making in autonomous systems or agents.

As part of the application »AI framework for autonomous systems«, a dependable driving assistant in is developed in order to demonstrate our technology. The typical processing pipeline of autonomous vehicles is made up from perception, routing, behavioral planning, motion planning and actuators. We focus on the implementation of behavioral planning by dependable reinforcement learning. Algorithms are developed that can steer the car in complex and critical situations by a safe policy. This policy is trained to be performant and efficient while adhering to strict safety constraints. Moreover, by extracting decision trees via imitation learning, the AI-driven behavior is made not only interpretable, but also verifiable. This showcases that DRL can be both dependable and efficient in critical use-cases.